This is a repository copy of Counting independent sets in graphs with bounded bipartite pathwidth. White Rose Research Online URL for this paper: http://eprints.whiterose.ac.uk/148669/ Version: Accepted Version Proceedings Paper: Dyer, M, Greenhill, C and Müller, H (2019) Counting independent sets in graphs with bounded bipartite pathwidth. In: Sau, I and Thilikos, D, (eds.) Graph-Theoretic concepts in computer science. Lecture Notes in Computer Science 11789. WG 2019, 19-21 Jun 2019, Vall de Nuria, Catalonia, Spain. Springer Verlag . ISBN 9783030307851 https://doi.org/10.1007/978-3-030-30786-8_23 © Springer Nature Switzerland AG 2019. This is an author produced version of a paper published in Lecture Notes in Computer Science. Uploaded in accordance with the publisher's self-archiving policy. Reuse Items deposited in White Rose Research Online are protected by copyright, with all rights reserved unless indicated otherwise. They may be downloaded and/or printed for private study, or other acts as permitted by national copyright laws. The publisher or other rights holders may allow further reproduction and re-use of the full text version. This is indicated by the licence information on the White Rose Research Online record for the item. Takedown If you consider content in White Rose Research Online to be in breach of UK law, please notify us by emailing [email protected] including the URL of the record and the reason for the withdrawal request. [email protected] https://eprints.whiterose.ac.uk/ Counting independent sets in graphs with bounded bipartite pathwidth Martin Dyer1, Catherine Greenhill2, and Haiko M¨uller1 1 School of Computing, University of Leeds, Leeds LS2 9JT, UK. Research supported by EPSRC grant EP/S016562/1. {M.E.Dyer|H.Muller}@leeds.ac.uk 2 School of Mathematics and Statistics, UNSW Sydney, NSW 2052, Australia. Research supported by Australian Research Council grant DP19010097. [email protected] Abstract. The Glauber dynamics can efficiently sample independent sets almost uniformly at random in polynomial time for graphs in a certain class. The class is determined by boundedness of a new graph parameter called bipartite pathwidth. This result, which we prove for the more general hardcore distribution with fugacity λ, can be viewed as a strong generalisation of Jerrum and Sinclair’s work on approximately counting matchings. The class of graphs with bounded bipartite path- width includes line graphs and claw-free graphs, which generalise line graphs. We consider two further generalisations of claw-free graphs and prove that these classes have bounded bipartite pathwidth. Keywords: Markov chain Monte Carlo algorithm · Fully polynomial- time randomized approximation scheme · Independent set · Pathwidth 1 Introduction We will show that we can approximate the number of independent sets in graphs for which all bipartite induced subgraphs are well structured, in a sense that we will define precisely. Our approach is to generalise the Markov chain analysis of Jerrum and Sinclair [19] for the corresponding problem of counting matchings. Their canonical path argument relied on the fact that the symmetric difference of two matchings of a given graph G is a bipartite subgraph of G consisting of a disjoint union of paths and even-length cycles. We introduce a new graph parameter, which we call bipartite pathwidth, to enable us to give the strongest generalisation of the approach of [19]. 1.1 Independent set problems For a given graph G, let I(G) be the set of all independent sets in G. The independence number α(G) = max{|I| : I ∈ I(G)} is the size of the largest independent set in G. The problem of finding α(G) is NP-hard in general, even in various restricted cases, such as degree-bounded graphs. However, polynomial 2 Martin Dyer, Catherine Greenhill, and Haiko M¨uller time algorithms have been constructed for finding a maximum independent set, for various graph classes. The most important case has been matchings, which are independent sets in the line graph L(G) of G. This has been generalised to larger classes of graphs, for example claw-free graphs [24], which include line graphs [4], and fork-free graphs [1], which include claw-free graphs. Counting independent sets in graphs is known to be #P-complete in gen- eral [26], and in various restricted cases [15, 30]. Exact counting is known only for some restricted graph classes. Even approximate counting is NP-hard in gen- eral, and is unlikely to be in polynomial time for bipartite graphs [11]. For some classes of graphs, for example line graphs, approximate counting is known to be possible [19, 20]. The most successful Markov chain approach relies on a close correspondence between approximate counting and sampling uni- formly at random [21]. It was applied to degree-bounded graphs in [23] and [12]. In his PhD thesis [22], Matthews used a Markov chain for sampling independent sets in claw-free graphs. His chain, and its analysis, generalises that of [19]. Several other approaches to approximate counting have been successfully applied to the independent set problem. Weitz [31] used the correlation decay approach on degree-bounded graphs, resulting in an FPTAS for counting inde- pendent sets in graphs with degree at most 5. Sly [29] gave a matching NP- hardness result. The correlation decay method was also applied to matchings in [3], and was extended to complex values of λ in [16]. Recently, Efthymiou et al. [14] proved that the Markov chain approach can (almost) produce the best results obtainable by other methods. The independence polynomial PG(λ) of a graph G is defined in (1) below. The Taylor series approach of Barvinok [2] was used by Patel and Regts [25] to give a FPTAS for PG(λ) in degree-bounded claw-free graphs. The success of the method depends on the location of the roots of the independence polynomial. Chudnovsky and Seymour [7] proved that all these roots are real, and hence they are all negative. Then the algorithm of [25] is valid for all complex λ which are not real and negative. In this extended abstract (for proofs see [13]), we return to the Markov chain approach. 1.2 Preliminaries We write [m] = {1, 2, . , m} for any positive integer m, and let A ⊕ B denote the symmetric difference of sets A, B. For graph theoretic definitions not given here, see [10]. Throughout this paper, all graphs are simple and undirected. G[S] denotes the subgraph of G induced by the set S and N(v) denotes the neighbourhood of vertex v. Given a graph G = (V,E), let Ik(G) be the set of independent sets of G of size k. The independence polynomial of G is the partition function α(G) |I| k PG(λ)= λ = Nk λ , (1) I∈IX(G) kX=0 Counting independent sets in graphs with bounded bipartite pathwidth 3 where Nk = |Ik(G)| for k = 0,...,α. Here λ ∈ C is called the fugacity. We consider only real λ and assume λ ≥ 1/n to avoid trivialities. We have N0 = 1, n N1 = n and Nk ≤ for k = 2,...,n. Thus it follows that for any λ ≥ 0, k α(G) n 1+ nλ ≤ P (λ) ≤ λk ≤ (1 + λ)n. (2) G k kX=0 Note also that PG(0) = 1 and PG(1) = |I(G)|. An almost uniform sampler for a probability distribution π on a state Ω is a randomised algorithm which takes as input a real number δ > 0 and outputs a sample from a distribution µ such that the total variation distance 1 2 x∈Ω |µ(x) − π(x)| is at most δ. The sampler is a fully polynomial almost uniformP sampler (FPAUS) if its running time is polynomial in the input size n and log(1/δ). The word “uniform” here is historical, as it was first used in the case where π is the uniform distribution. We use it in a more general setting. If w : Ω → R is a weight function, then the Gibbs distribution π satisfies π(x)= w(x)/W for all x ∈ Ω, where W = x∈Ω w(x). If w(x) = 1 for all x ∈ Ω then π is uniform. For independent sets withP w(I)= λ|I|, we have |I| π(I)= λ /PG(λ), (3) and is often called the hardcore distribution. Jerrum, Valiant and Vazirani [21] showed that approximating W is equivalent to the existence of an FPAUS for π, provided the problem is self-reducible. Counting independent sets in a graph is a self-reducible problem. (2) can be tightened to α α α n (nλ)k 1 P (λ) ≤ λk ≤ ≤ (nλ)α ≤ e(nλ)α. (4) G k k! k! kX=0 kX=0 kX=0 2 Markov chains 2.1 Mixing time For general information on Markov chains and approximate counting see [18]. Consider a Markov chain on state space Ω with stationary distribution π and transition matrix P. Let pn be the distribution of the chain after n steps. We will assume that p0 is the distribution which assigns probability 1 to a fixed initial state x ∈ Ω. The mixing time of the Markov chain, from initial state x ∈ Ω, is τx(ε) = min{n : dTV(pn, π) ≤ ε}, where dTV(pn, π) is the total variation distance between pn and π. In the case of the Glauber dynamics for independent sets, the stationary distribution π satisfies (3), and in particular −1 π(∅) = PG(λ). We will always use ∅ as our starting state. Let βmax = max{β1, |β|Ω|−1|}, where β1 is the second-largest eigenvalue and β|Ω|−1 is the smallest eigenvalue of P.